Comparison of HP Filter and the Hamilton’s Regression

Abstract

In this paper we examine if the use of Hamilton’s regression filter significantly modifies the cyclical components concerning unemployment in Greece compared with those using the Hodrick–Prescott double filter (HP). Hamilton suggested the use of a regression filter in order to overcome some of the drawbacks of the HP filter, which contains the presence of false cycles, the bias in the end of the sample, and the ad-hoc assumptions for the parameters’ smoothing. Thus, our paper examines two widely used detrending methods for the extraction of cyclical components, including techniques of deterministic detrending as well as stochastic detrending. Using quarterly data for the unemployment of Greece in a macroeconomic model decomposition, we indicate that trend components and cycle components of Hamilton’s filter regression led to significantly larger cycle volatilities than those from the HP filter. The dynamic forecasting in the sample, occurred both with autoregressive forecasting, that yields steady forecasts for a wide variety of non-stationary procedures, and with the HP filter, along with its constraints at the end of the time series. The results of the paper showed that the dynamic forecasting of the HP filter is better than that of Hamilton’s in all assessment measures.